Leo Parker DiracFounder

United States - Seattle, Washington

Slamming my foot down on the accelerator pedal of technological development.

Helping accelerate the adoption of AI to solve society's problems. Pushing the second or third derivative of progress.

Please don't message me here if you actually want me to read your note -- signal/noise ratio way too low. Visit http://leodirac.com/contact for instructions.

Since 2012 I've devoted myself to understanding and deploying deep learning, having understood its potential from as soon as AlexNet won the ImageNet competition. At Amazon I spent many years building world-class infrastructure to enable machine learning in general, understanding the ML+Engineering systems interactions as well as anybody anywhere. On ML science, I read a lot of papers and publish little. I maintain research-level expertise in AutoML (Bayesian optimization, meta-learning, hyperparameter tuning, neural architecture search, etc.) and neural network training (non-convex optimization, geometry of loss surfaces, distributed training algorithms). I also have a working practical working knowledge of reinforcement learning, computer vision, natural language processing, active learning, representation learning, and adversarial techniques. I tend to dive into topics that I think have high leverage to improve the world.